Research On Cloud Computing Load Detection And Balancing Technology

With the development of the times,the basic resources of computing are also striding forward.In the early computing resource allocation,IT resource is a host-based scheduling unit,and most of users'applications run directly in bare computers.This pattern highlights more and more problems in today's context.For example,when the host has an unexpected downtime,user misoperation leads to the failure of all applications,and the inefficiency of resource utilization.IT architecture is turning to virtualization,but it also faces some new problems,such as the difficulty of virtualization resource life cycle management,the need for administrators to have a certain virtual knowledge reserve,the difficulty of monitoring global virtualization status,the high availability of virtual machines can not be guaranteed,and so on.In order to solve the above problems,this design provides a design idea and partial implementation of virtualization(private cloud)management architecture,aiming at solving the problems of automatic deployment of virtualization resources,life cycle management,performance monitoring and high availability of virtualization resources.The architecture is divided into several parts:virtual resource status detection,automatic scheduling,performance acquisition and automatic failover.The above main functional modules are subdivided into many atomic level functions.It can achieve real-time monitoring of virtualized resources,active and passive performance collection in various ways,and active handling of local faults.Finally,through coupling these functional modules,the ultimate goal of the framework is to achieve full automatic operation without manual intervention on the premise of providing artificial intervention interface.This design is based on Python as a programming language,using MySQL as the back-end database,using three-terminal(front-end,back-end,proxy)mode to design.The coupling problem of state detection and scheduling in virtual environment is solved.Provide pluggable algorithm module interface to dynamically adapt to different management requirements.It achieves automatic scheduling based on history and current state without manual intervention.In addition,this design provides a good interface for managers to manage,and provides a basic available model and preliminary implementation for large-scale virtualization applications.After testing,the framework can achieve the expected goal of virtualization of fault detection,high availability,and scheduling functions.